Most long-term memories are forgotten, becoming progressively less likely to be recalled. Still, some memory fragments may persist, as savings memory (easier relearning) can be detected long after recall has become impossible. What happens to a memory trace during forgetting that makes it inaccessible for recall and yet still effective to spark easier re-learning? We are addressing this question by tracking the transcriptional changes that accompany learning and then forgetting of a long-term sensitization memory in the tail-elicited siphon withdrawal reflex of Aplysia californica. First, we tracked savings memory. We found that even though recall of sensitization fades completely within 1 week of training, savings memory is still detectable at 2 weeks post training. Next, we tracked the time-course of regulation of 11 transcripts we previously identified as potentially being regulated after recall has become impossible. Remarkably, 3 transcripts still show strong regulation 2 weeks after training and an additional 4 are regulated for at least 1 week. These long-lasting changes in gene expression always begin early in the memory process, within 1 day of training. We present a synthesis of our results tracking gene expression changes accompanying sensitization and provide a testable model of how sensitization memory is forgotten.